我试图将2个数据帧列合并为1,但是当我尝试根据特定大小进行操作时,第二个数据帧列无法正确复制。
我尝试了下面粘贴的以下代码。
import pandas as pd
def readDataFile():
fileName = "year.csv"
dateparse = lambda x: pd.datetime.strptime(x, '%Y-%m-%d %H:%M:%S')
dfY = pd.read_csv(fileName, parse_dates=['date'], date_parser=dateparse)
fileName = "month.csv"
dateparse = lambda x: pd.datetime.strptime(x, '%Y-%m-%d %H:%M:%S')
dfM = pd.read_csv(fileName, parse_dates=['date'], date_parser=dateparse)
newDF = pd.DataFrame()
newDF['date_y'] = dfY['date']
newDF['year_y_n'] = dfY['Y_N']
newDF['date_m'] = dfM['date'][len(dfM) - len(dfY):len(dfM)]
newDF['year_y_n'] = dfM['Y_N'][len(dfM) - len(dfY):len(dfM)]
print newDF
readDataFile()
文件:month.csv
date,Y_N
2018-03-14 04:00:00,N
2018-04-03 04:00:00,N
2018-05-31 04:00:00,Y
2018-06-14 04:00:00,N
2018-07-30 04:00:00,N
2018-08-31 04:00:00,Y
2018-09-28 04:00:00,N
2018-10-10 04:00:00,N
2018-11-07 04:00:00,Y
2018-12-31 04:00:00,N
2019-01-31 04:00:00,N
2019-02-05 04:00:00,Y
2019-03-29 04:00:00,N
2019-04-30 04:00:00,Y
2019-05-03 04:00:00,N
2019-06-03 04:00:00,Y
文件:year.csv
date,Y_N
2014-05-23 04:00:00,Y
2015-12-21 04:00:00,N
2016-05-03 04:00:00,Y
2017-12-20 04:00:00,N
2018-06-14 04:00:00,N
2019-06-25 04:00:00,N
以下是当前结果:
date_y year_y_n date_m month_y_n
0 2014-05-23 04:00:00 Y NaT NaN
1 2015-12-21 04:00:00 N NaT NaN
2 2016-05-03 04:00:00 Y NaT NaN
3 2017-12-20 04:00:00 N NaT NaN
4 2018-06-14 04:00:00 N NaT NaN
5 2019-06-25 04:00:00 N NaT NaN
预期结果是:
date_y year_y_n date_m month_y_n
2014-05-23 04:00:00 Y 2019-01-31 04:00:00 N
2015-12-21 04:00:00 N 2019-02-05 04:00:00 Y
2016-05-03 04:00:00 Y 2019-03-29 04:00:00 N
2017-12-20 04:00:00 N 2019-04-30 04:00:00 Y
2018-06-14 04:00:00 N 2019-05-03 04:00:00 N
2019-06-25 04:00:00 N 2019-06-03 04:00:00 Y
答案 0 :(得分:0)
假设您有任意数量的数据帧dfA
,dfB
,dfC
等。您想合并它们,但是它们的大小不同。最基本的方法是将它们连接起来:
df = pd.concat([dfA, dfB, dfC], axis=1)
但是,如果数据帧的大小不同,则会丢失行。如果您不关心保留哪些行,则可以删除缺少值的行:
df.dropna()
但是,如果您特别想使用每个数据帧的最后 N 行,其中 N 是最小数据帧的长度,则您需要做更多的工作。但是我会等一下,看看那是不是你想要的。
旧答案:
合并可能比这简单得多。使用pd.merge
:
pd.merge(dfY, dfM[-len(dfY):].reset_index(),
suffixes=['_y', '_m'], left_index=True, right_index=True)
dfM[-len(dfY):]
获取dfM
的最后 N 行,其中 N 是dfY
的长度。 .reset_index()
使dfM
的子集的索引从0开始,因此它可以正确地与dfY
对齐。suffixes=['_y', '_m']
使列名保持不同。您可以根据需要将其重命名。答案 1 :(得分:0)
问题与索引有关。 如果您运行以下代码:
newDF = pd.DataFrame()
newDF['date_y'] = dfY['date']
print(newDF)
您将获得输出:
date_y
0 2014-05-23 04:00:00
1 2015-12-21 04:00:00
2 2016-05-03 04:00:00
3 2017-12-20 04:00:00
4 2018-06-14 04:00:00
5 2019-06-25 04:00:00
索引从0开始
运行此:
newDF = pd.DataFrame()
newDF['date_m'] = dfM['date'][len(dfM) - len(dfY):len(dfM)]
print(newDF)
您将获得输出:
date_m
10 2019-01-31 04:00:00
11 2019-02-05 04:00:00
12 2019-03-29 04:00:00
13 2019-04-30 04:00:00
14 2019-05-03 04:00:00
15 2019-06-03 04:00:00
此处,索引从10开始
因此,您需要重置dfM数据帧的“日期”和“ Y_N”列的索引,如下所示:
def readDataFile():
fileName = "year.csv"
dateparse = lambda x: pd.datetime.strptime(x, '%Y-%m-%d %H:%M:%S')
dfY = pd.read_csv(fileName, parse_dates=['date'], date_parser=dateparse)
fileName = "month.csv"
dateparse = lambda x: pd.datetime.strptime(x, '%Y-%m-%d %H:%M:%S')
dfM = pd.read_csv(fileName, parse_dates=['date'], date_parser=dateparse)
newDF = pd.DataFrame()
newDF['date_y'] = dfY['date']
newDF['year_y_n'] = dfY['Y_N']
# Changes made on this line.
newDF['date_m'] = dfM['date'][len(dfM) - len(dfY):len(dfM)].reset_index(drop=True)
newDF['month_y_n'] = dfM['Y_N'][len(dfM) - len(dfY):len(dfM)].reset_index(drop=True)
print(newDF)
readDataFile()
输出:
date_y year_y_n date_m month_y_n
0 2014-05-23 04:00:00 Y 2019-01-31 04:00:00 N
1 2015-12-21 04:00:00 N 2019-02-05 04:00:00 Y
2 2016-05-03 04:00:00 Y 2019-03-29 04:00:00 N
3 2017-12-20 04:00:00 N 2019-04-30 04:00:00 Y
4 2018-06-14 04:00:00 N 2019-05-03 04:00:00 N
5 2019-06-25 04:00:00 N 2019-06-03 04:00:00 Y